1521-0111/93/6/619–630$35.00 https://doi.org/10.1124/mol.118.112086MOLECULAR PHARMACOLOGY Mol Pharmacol 93:619–630, June 2018Copyright ª 2018 by The American Society for Pharmacology and Experimental Therapeutics

Identification of Global and Ligand-Specific Calcium SensingReceptor Activation Mechanisms s

Andrew N. Keller, Irina Kufareva, Tracy M. Josephs, Jiayin Diao, Vyvyan T. Mai,Arthur D. Conigrave, Arthur Christopoulos, Karen J. Gregory, and Katie LeachDrug Discovery Biology, Monash Institute of Pharmaceutical Sciences and Department of Pharmacology, Monash University,Parkville, Victoria, Australia (A.N.K., T.M.J., J.D., V.T.M., A.C., K.J.G., K.L.); Skaggs School of Pharmacy and PharmaceuticalSciences, University of California, La Jolla, San Diego, California (I.K.); and School of Life and Environmental Sciences, CharlesPerkins Centre (D17), University of Sydney, New South Wales, Australia (A.D.C.)

Received February 7, 2018; accepted April 6, 2018

ABSTRACTCalcium sensing receptor (CaSR) positive allosteric modulators(PAMs) are therapeutically important. However, few are approvedfor clinical use, in part due to complexities in assessing allostery at areceptor where the endogenous agonist (extracellular calcium) ispresent in all biologic fluids. Such complexity impedes efforts toquantify and optimize allosteric drug parameters (affinity, coopera-tivity, and efficacy) that dictate PAM structure-activity relationships(SARs). Furthermore, an underappreciation of the structural mecha-nisms underlying CaSR activation hinders predictions of how PAMSAR relates to in vitro and in vivo activity. Herein, we combined site-directedmutagenesis andcalciummobilizationassayswith analyticalpharmacology to compare modes of PAM binding, positive modu-lation, and agonism. We demonstrate that 3-(2-chlorophenyl)-N-((1R)-1-(3-methoxyphenyl)ethyl)-1-propanamine (NPSR568) binds toa 7 transmembrane domain (7TM) cavity common to class C G

protein-coupled receptors and used by (aR)-(2)-a-methyl-N-[3-[3-[trifluoromethylphenyl]propyl]-1-napthalenemethanamine (cina-calcet) and 1-benzothiazol-2-yl-1-(2,4-dimethylphenyl)-ethanol(AC265347); however, there are subtle distinctions in the contributionof select residues to the binding and transmission of cooperativity byPAMs. Furthermore, we reveal some common activation mecha-nisms used by different CaSR activators, but also demon-strate some differential contributions of residues within the7TM bundle and extracellular loops to the efficacy of thePAM-agonist, AC265347, versus cooperativity. Finally, weshow that PAMS potentiate the affinity of divalent cations.Our results support the existence of both global and ligand-specific CaSR activation mechanisms and reveal that allo-steric agonism is mediated in part via distinct mechanisms topositive modulation.

IntroductionThe human calcium sensing receptor (CaSR) is a class C G

protein-coupled receptor (GPCR) that negatively regulates para-thyroid hormone secretion in response to serum extracellularcalcium (Ca21o ) concentrations. As such, the small-moleculepositive allosteric modulator (PAM), (aR)-(2)-a-methyl-N-[3-[3-[trifluoromethylphenyl]propyl]-1-napthalenemethanamine(cinacalcet), is approved for the treatment of secondary hyper-parathyroidism, and some instances for the treatment of primaryhyperparathyroidism (Nemeth and Goodman, 2016). Cinacalcet

has also been used off-label to correct disorders caused bynaturally occurring loss-of-function CaSR mutations (Mayret al., 2016). Although various CaSR small-molecule PAMchemotypes have been discovered, cinacalcet remains the onlyone on themarket. Failure to translate newerCaSRPAMs to theclinic may reflect complexities associated with detecting andquantifying in vitro allostericmodulator effects at classCGPCRs(Leach and Gregory, 2017), thus hampering structure-activityrelationship (SAR) optimization efforts and predictions of in vivodrug efficacy, and impeding dissection of allosteric mechanismsof action. For instance, despite cinacalcet being the first clinicallyapproved GPCR allosteric modulator to reach the market,surprisingly little is known about how cinacalcet and otherPAMs mediate positive allosteric modulation and/or agonism atthe CaSR. Understanding GPCR activation mechanisms canhelp guide SAR optimization of ligands with desired activity.Structural studies of class A and B GPCRs indicate that

during receptor activation a rotational and vertical movementof transmembrane (TM) 6 and TM7 relative to TM3 occurs,

K.L. and K.J.G. are Australian Research Council Future Fellowshiprecipients. A.C. is a Senior Principal Research Fellow of the AustralianNational Health and Medical Research Council.

This work was funded by the Australian National Health and MedicalResearch Council [Grants APP1026962 and APP1085143] and the NationalInstitutes of Health National Institute of General Medical Sciences andNational Institute of Allergy and Infectious Diseases [Grants GM117424 andAI118985, respectively].

https://doi.org/10.1124/mol.118.112086.s This article has supplemental material available at molpharm.

aspetjournals.org.

ABBREVIATIONS: AC265347, 1-benzothiazol-2-yl-1-(2,4-dimethylphenyl)-ethanol; ANOVA, analysis of variance; Ca21i , intracellular calcium; Ca21o ,extracellular calcium; CaSR, calcium sensing receptor; cinacalcet, (aR)-(2)-a-methyl-N-[3-[3-[trifluoromethylphenyl]propyl]-1-napthalenemethanamine;ECL, extracellular loop; GPCR, G protein-coupled receptor; NPS R568, 3-(2-chlorophenyl)-N-((1R)-1-(3-methoxyphenyl)ethyl)-1-propanamine; PAM,positive allosteric modulator; SAR, structure-activity relationship; TM, transmembrane; 7TM, 7 transmembrane domain; VFT, Venus flytrap; WT, wildtype.

619

http://molpharm.aspetjournals.org/content/suppl/2018/04/10/mol.118.112086.DC1Supplemental material to this article can be found at:

at ASPE

T Journals on July 8, 2020

molpharm

.aspetjournals.orgD

ownloaded from

resulting in conformational changes at the intracellular face ofthe receptor that promote transducer coupling (reviewed in deGraaf et al., 2017; Manglik and Kruse, 2017). For class CGPCRs, most endogenous agonists predominantly bind in thelarge extracellular Venus flytrap (VFT) domain, stabilizingactivating conformational rearrangements that transmitthrough the 7 transmembrane domain (7TM) spanning re-gions and/or extracellular loops (ECLs) to the intracellularside of the receptor. Many synthetic molecules that targetclass C GPCRs bind to allosteric sites in the 7TM domainand/or ECL regions. Furthermore, for the CaSR, the 7TMand/or ECL domains also contain additional binding sites forendogenous agonists including polyamines and Ca21o (Ray andNorthup, 2002; Leach et al., 2016). However, the 7TMmolecular mechanisms that underlie class C GPCR allosteric(and indeed orthosteric) activation are largely unknown. Thisis due to a complete lack of active-state class C GPCR 7TMstructures, but also, in part, due to the difficulties ininterpreting mutation-based structure-function studies to de-lineate residue contributions to the three key molecularproperties that govern allosteric activity; ligand affinity,cooperativity, and efficacy.In light of the difficulties associated with studying GPCR

allosteric modulator effects, we developed pharmacologicaland analytical methods to robustly quantify allosteric modu-lation of GPCRs (Leach et al., 2007), and recently employedthese methods to probe class C GPCR allostery (Gregory et al.,2012, 2013, 2014; Leach et al., 2016). We demonstrated

unprecedented insights into class C GPCR drug actionsby combining pharmacological, analytical, and detailedstructure-function analyses. For instance, we identified aCaSR 7TM domain cavity that accommodates small-molecule CaSR PAMs and negative allosteric modulators,and showed that structurally diverse modulators bind todistinct regions in this cavity; the arylalkylamine PAM,cinacalcet, binds toward the top of the 7TM domain cavity,whereas the structurally distinct CaSR PAM-agonist,1-benzothiazol-2-yl-1-(2,4-dimethylphenyl)-ethanol(AC265347), binds deeper in the pocket (Leach et al., 2016).Furthermore, we mapped residues that contribute to PAMpotentiation of the CaSR’s endogenous agonist, Ca21o , re-vealing both shared and distinct activation mechanisms usedby cinacalcet and AC265347 (Leach et al., 2016).Given the importance of understanding the structural basis

underlying ligand binding and activation mechanisms at class CGPCRs, the current study sought to build on our prior analysis ofCaSR allostery by examining the structural basis of positiveallosteric modulation mediated by the small-molecule PAM, 3-(2-chlorophenyl)-N-((1R)-1-(3-methoxyphenyl)ethyl)-1-propanamine(NPS R568). The binding of NPS R568 and other arylalkylaminePAMs is predicted to overlap with the cinacalcet binding site.Given that arylalkylamine PAMs display comparable in vitropharmacology, we hypothesized that NPS R568 would demon-strate similar structure-function relationships to cinacalcet. Wealso hypothesized that the PAM-agonist, AC265347, mediatesagonism via engagement of residues that transmit positive

Fig. 1. Studying positive allosteric modulation at the CaSR. (A) Structures of the arylalkylamine PAM, NPS R568, and the PAM-agonist, AC265347,investigated in this study. (B) According to the operational model of allosterism, the CaSR PAM effects on orthosteric agonist signaling are governed byPAM affinity (KB), binding and/or efficacy cooperativity between the PAM and the orthosteric agonist (a and b, respectively), and the efficacy of the PAM(tB); ab governs the magnitude of agonist potentiation, reflected by an increase in agonistEmax and/or potency, and enhanced vehicle-mediated signalingplus a reduction in the slope of the transducer function, n, if ambient orthosteric agonists are present in the buffer. If the PAM is also an agonist, tB willdictate PAM-mediated receptor signaling in the presence of vehicle. (C) Snake diagram of the primary CaSR 7TM domain and ECL sequence, indicatingresidues that were mutated in the current study. Mutations at residues previously shown to be important for the binding of the PAMs, cinacalcet, and/orAC265347, their cooperativity with Ca2+o , or both (Leach et al., 2016), are shown in blue, red, and orange, respectively. All engineered mutations weresubstituted with alanine, with the exception of A6151.42, A7725.39, A8407.35, and A8447.3, which were substituted with valine, and E8377.32, which wasmutated to aspartic acid and isoleucine. The residue highlighted in green indicates where a naturally occurringmutation alters the activity of somePAMs(Cook et al., 2015). Arrows point to the �.50 conserved class C amino acid residues based on a modified Ballesteros-Weinstein numbering system (Doréet al., 2014).

620 Keller et al.

at ASPE

T Journals on July 8, 2020

molpharm

.aspetjournals.orgD

ownloaded from

modulation. Thus, we usedmutagenesis to comparemechanismsof positive allosteric modulation and direct allosteric agonismmediated byNPSR568 andAC265347 (Fig. 1A).We identify bothglobal and ligand-specific activation mechanisms at the CaSR,which could inform future drug discovery efforts seeking torationally optimize the activity of CaSR activators and/orpotentiators.

Materials and MethodsMaterials. Dulbecco’smodified Eagle’smedium, blasticidin SHCl,

and fetal bovine serumwere obtained from Invitrogen (Carlsbad, CA),while hygromycin B was obtained from Roche (Mannheim, Germany).The Flp-In TREx human embryonic kidney 293 cells and Fluo-4 AMacetoxymethyl ester were purchased from Invitrogen. Quikchangemutagenesis kits were purchased from Agilent Technologies (SantaClara, CA). NPS R568 was prepared as described previously (Daveyet al., 2012). AC265347 and all other chemicals were from SigmaAldrich (St Louis, MO).

Generation andMaintenance of Cell Lines Expressing Wild-Type (WT) and Mutant CaSRs. The generation of DNA and cellsstably expressing c-myc-tagged WT and mutant human CaSRs inpcDNA5/frt/TO have been described previously (Davey et al., 2012;Leach et al., 2016). Cells were maintained in Dulbecco’s modifiedEagle’s medium containing 5%–10% fetal bovine serum, 200 mg/mlhygromycin B, and 5 mg/ml blasticidin S HCl.

Intracellular Calcium (Ca21i ) Mobilization Assays. Cells were

seeded in clear 96-well plates coated with poly-D-lysine (50 mg/ml21)at 80,000 cells/well and incubated overnight in the presence of100 ng/ml21 tetracycline. The following day, cells were washed withassay buffer (150 mM NaCl, 2.6 mM KCl, 1.18 mM MgCl2, 10 mMD-Glucose, 10 mM HEPES, 0.1 mM CaCl2, 0.5% bovine serumalbumin, and 4 mM probenecid at pH 7.4) and loaded with Fluo-4AM acetoxymethyl ester (1 mM in assay buffer) for 1 hour at 37°C.Cells were washed with assay buffer prior to the addition of freshassay buffer. For interaction studies between Ca21o and PAMs,modulators were coadded with Ca21o . For assays in the absence ofambient divalent cations, cells were treated as described previously,except that the final assay buffer and the buffer used for washing andloading the cells did not contain CaCl2 or MgCl2.

For all studies, each well was treated with a single agonist and/ormodulator concentration. The release of Ca21i was measured at 37°Cusing Flexstation 1 or 3 (Molecular Devices, Sunnyvale, CA). Fluo-rescence was detected for 60 seconds at 485 nm excitation and 525 nmemission, and the peak Ca21i mobilization response (approximately12 seconds after agonist addition) was used for the subsequentdetermination of the agonist response. Relative peak fluorescenceunits were normalized to the fluorescence stimulated by 1 mMionomycin to account for differences in cell number and loadingefficiency.

Data Analysis. All nonlinear regression analyses were performedusing GraphPad Prism 7 (GraphPad Software, San Diego, CA).Parametric measures of potency, affinity, and cooperativity wereestimated as logarithms (Christopoulos, 1998). Interaction experi-ments between Ca21o and allosteric modulators were fitted to anoperational model of allosterism (Leach et al., 2007; Aurelio et al.,2009; Leach et al., 2010), which was modified to accommodate thepresence of ambient agonist (Davey et al., 2012):

E5ðEmax 2 basalÞ ½ð½A�1 ½Ae�Þ ðKB 1ab½B�Þ 1 tB ½B� ½EC50��n

½EC50� ðKB 1 ½B�Þ 1 ½ð½A�1 ½Ae�Þ ðKB 1 ab ½B�Þ 1 tB ½B� ½EC50��n(1)

where [A] denotes the concentration of exogenously added orthostericagonist to stimulate the response; [Ae] is the ambient orthostericagonist concentration in the buffer; [B] is the molar allosteric ligandconcentration; KB is the functional affinity of the allosteric ligand; tB

denotes allosteric ligand intrinsic efficacy; a and b are the allostericeffects on orthosteric ligand binding affinity and efficacy, respectively(estimated as a composite ab value); Emax is the maximal possiblesystem response; and n is the slope factor of the transducer function.

Concentration-response curves for Ca21o , NPS R568, and AC265347in the absence of ambient divalent cations were fitted to the followingform of an operational model of agonism to obtain estimates of PAMaffinity and efficacy:

E5

�ðEmax 2basalÞ tB ½B�ðKB 1 ½B�Þ1 tB ½B�

�n

(2)

where all parameters are as described for eq. 1.Statistics. An F test (P , 0.001) was used to determine whether

eq. 1 best fit interaction data when NPS R568 and AC265347 wereassumed to have no agonist efficacy at theWT andmutant CaSRs (i.e.,whether the best-fit value for tB was equal to or differed from zero).Statistical differences between the Ca21o pEC50, or PAM pKB andlogab at theWT versus the mutant receptors were determined by one-way analysis of variance (ANOVA) with Dunnett’s multiple compar-isons post-test, where significance was defined as P, 0.05. Statisticaldifferences between PAM pKB at the WT and mutant receptors in theabsence versus the presence of ambient divalent cations was de-termined by two-way ANOVA with Sidak’s comparisons post-test,where significance was considered as P , 0.05.

Molecular Modeling and Docking Studies. A three-dimensional model of the CaSR 7TM domain was constructed byhomology with mGlu5 (PDB: 5CGD) (Christopher et al., 2015) usingthe ICM software package (Molsoft, San Diego, CA) (Abagyanand Totrov, 1994) as previously described (Leach et al., 2016). The7TM helices in the model were additionally refined to accom-modate irregularities such as nonconserved proline residues (e.g.,Pro6823.34) and a one-residue insertion in the extracellular part ofTM5 (Ile7775.44). Compound docking relied on the pharmacophoreproperties of the TM cavity as well as the results of earlier and presentmutagenesis studies indicating that mutation of Glu8377.32 abrogatesactivity of all arylalkylamines. A salt bridge is predicted to occurbetween the protonated arylalkylamine secondary amine and theGlu8377.32 side chain (Petrel et al., 2003, 2004; Miedlich et al., 2004;Bu et al., 2008; Leach et al., 2016); therefore, the formation of this saltbridge was enforced in our docking studies by imposing a harmonicdistance restraint between the corresponding atoms. The arylalkyl-amine docking proceeded by extensive conformational sampling of theligands and the residue side chains lining the pocket in the internalcoordinates in the presence of this distance restraint. All complexeswere further refined by local minimization in the presence of distancerestraints maintaining receptor secondary and tertiary structures,and inspected manually.

ResultsRationale for the Study. We have previously evaluated

the effects of CaSR 7TM and ECL mutations on allostericmodulator activity using an interaction paradigm (Daveyet al., 2012; Cook et al., 2015; Leach et al., 2016). Here, theeffect of a CaSR PAM can manifest as: 1) modulation of theexogenously added agonist to stimulate the response, reflectedby a change in the added agonist potency and/or Emax; 2)modulation of ambient agonist present in the buffer, reflectedby an increase in the baseline response and/or a shallowing ofthe agonist concentration-response curve; and 3) direct PAM-mediated agonism, also reflected by an increase in thebaseline response (Davey et al., 2012; Cook et al., 2015;Leach et al., 2016) (Fig. 1B). In this work, we have extendedour analysis to account for ambient Ca21o ions present in theassay buffer. Including the ambient Ca21o concentration in the

CaSR Activation Networks 621

at ASPE

T Journals on July 8, 2020

molpharm

.aspetjournals.orgD

ownloaded from

operational model of allosterism allows delineation of PAM-mediated increases in the vehicle (buffer) response due toagonism versus potentiation of ambient Ca21o (eq. 1). Accord-ingly, we are now able to quantify and delineate the effect ofCaSR mutations on PAM affinity (pKB), allosteric coopera-tivity with the orthosteric agonist (ab), and allosteric agonism(tB).Using this analytical paradigm, we examined two tool

compounds; NPS R568 and AC265347. NPS R568 (Fig. 1A)was among the first identified CaSR PAMs (Nemeth et al.,1998) and a precursor to cinacalcet, which is an approvedclinical treatment of human hyperparathyroidism. BothPAMs display similar in vitro pharmacology (Cook et al.,2015), despite subtle differences in their chemical structure;although NPS R568 and cinacalcet share a three carbonaliphatic chain linked to a secondary amine, they differ intheir terminal aromatics: chlorophenyl and methoxyphenylgroups in NPS R568 were replaced with tri-fluoromethyl-phenyl and naphthyl groups in cinacalcet. With these differ-ences in mind, we aimed to examine the SAR of NPS R568 incomparison with cinacalcet in more detail. Furthermore,whereas cinacalcet and NPS R568 exhibit little agonism, thechemically distinct CaSR PAM, AC265347, is both a PAM andan agonist (a PAM-agonist) (Cook et al., 2015). We were,therefore, interested in elucidating the different activation

mechanisms used by these chemically and pharmacologicallydistinct CaSR activators. Of further interest, AC265347exhibits biased CaSR modulation when compared with cina-calcet, presumably via stabilizing different active states(Davey et al., 2012; Leach et al., 2013, 2016; Cook et al., 2015).Therefore, in this work, we used mutagenesis in combina-

tion with an operational model of allosterism that explicitlyaccommodates ambient agonist levels (eq. 1), to characterizeimportant features of NPS R568 binding and positive cooper-ativity, and to directly probe AC265347 agonism. 7TMand ECL mutations were selected based on those known toalter the affinity and/or cooperativity of cinacalcet and/orAC265347 (Fig. 1C).NPS R568 Binds to a Common 7TM Allosteric

Binding Site. We first confirmed that NPS R568 boundwithin the same 7TM cavity used by other arylalkylaminemodulators (Leach et al., 2016). We fitted NPS R568 in-teraction data at the WT and 7TM mutant receptors to anoperational model of allosterism with ambient agonist toassess the impact of 7TM amino acid substitutions on NPSR568 affinity (pKB). WT and representative mutants areshown in Fig. 2 that exemplify the different profiles observed.Similar to our previous observations with cinacalcet (Leachet al., 2016), F6682.56A, F6843.36A, F6883.40A, E8377.32I, andI8417.36A mutations abolished NPS R568 activity or reduced

Fig. 2. 7TMmutations alter the activity of NPSR568 at theCaSR. Concentration-response curves to Ca2+o in the ab-sence and presence of NPS R568 were determined in Ca2+imobilization assays to identify mutations that altered NPSR568 activity (A–E). Representative mutants are shownthat reduced NPS R568 affinity (F6843.36A), increased NPSR568 affinity (A8447.39V), reduced NPS R568 cooperativity(V8176.49A), or increased NPS R568 affinity while decreas-ing cooperativity (F8216.53A). Data are mean + S.E.M.pooled from at least four separate experiments performedin duplicate. Curves are the best fit of the data to eq. 1.Vehicle (buffer) is plotted as the smallest [Ca2+o ] in accor-dance with the logarithmic scale.

622 Keller et al.

at ASPE

T Journals on July 8, 2020

molpharm

.aspetjournals.orgD

ownloaded from

NPS R568 pKB, whereas F8216.53A, E8377.32D, and A8447.39VincreasedNPS R568 pKB (Fig. 2; Fig. 3A; Supplemental Fig. 1;Table 1). Furthermore, although the magnitude of NPS R568potentiation of Ca21o potency at the Y8256.57A mutant wasreduced (Supplemental Fig. 1), a significant and maximal ∼2-fold increase in Ca21o potency was reached in the presence of0.1 mMNPSR568, suggesting the PAM pKB was increased. Toconfirm this finding, we performed interaction studies be-tween AC265347 and NPS R568. We found that 0.1–10 mMNPS R568 caused a dextral translocation of the AC265347concentration-response curve at the Y8256.42A mutant, withno apparent decrease in AC265347 Emax (Supplemental Fig.2); although limitations in AC265347 solubility preventeddetermination of complete AC265347 concentration-responsecurves in the presence of the higher NPS R568 concentrations.Nonetheless, these findings suggest that the interactionbetween NPS R568 and AC265347 is competitive, or anallosteric interaction governed by extremely high negativecooperativity. In support of a competitive or highly negatively

cooperative interaction, when AC265347 and NPS R568interactions at the Y8256.42A mutant were fitted to a modifiedHill/Gaddum/Schild equation (Motulsky and Christopoulos,2004; Langmead et al., 2006), the Schild slope was not signif-icantly different from unity. Under these circumstances, theconcentration of NPS R568 that shifts the AC265347 EC50

2-fold is equal to the NPS R568 pKB; this analysis confirmedthat the NPS R568 pKB was increased by the Y8256.57Amutation (Table 1). A similar increase in cinacalcet pKB hasbeen observed at this mutant (Leach et al., 2016). Our findingsare consistent with NPS R568 binding in the 7TM cavity in asimilar pose to cinacalcet, concordant with previous predictionsthat E8377.32 forms a hydrogen bond with the core secondaryamine in arylalkylamine modulators (Petrel et al., 2003, 2004;Miedlich et al., 2004; Bu et al., 2008; Leach et al., 2016). Withthis interaction, the terminal naphthyl (cinacalcet) or methox-yphenyl (NPS R568) moiety extends downward, p-stackingwith F6843.36 and forming hydrophobic interactions withF6682.56, F6883.40, and I8417.36.

Fig. 3. 7TM and ECLmutations alter the affinity and cooperativity of NPS R568 at the CaSR. (A) Concentration-response curves to Ca2+o in the absenceand presence of NPS R568 determined in Ca2+i mobilization assays were fitted to an operational model of allosterism (eq. 1) to determine the change inaffinity (DpKB) (A) and cooperativity (Dlogab) (C) of NPS R568 at the mutant CaSRs in comparison with the WT receptor. White bars represent nosignificant change in pKB or logab. Colored bars that sit above and below zero represent a significant increase or decrease in pKB or logab, respectively. A1.5-fold or lower reduction in pKB and/or logab is represented by yellow bars, a 1.6-fold to 2.5-fold reduction is represented by orange bars, and a greaterthan 2.5-fold reduction or increase is represented by red and blue bars, respectively. NA, negligible PAM activity. Data are the DpKB or Dlogab + S.E.M.calculated fromWT and mutant pKB or logab values and experiment numbers shown in Table 1; *denotes significant difference in comparison with WT(P, 0.05, one-way ANOVAwith Dunnett’s multiple comparisons post-test). Affinity (B) and cooperativity (D) residues are shown on a CaSR 7TMdomainhomology model (gray cartoon), along with the predicted NPS R568 pose from docking studies. Residue positions are shown as ball and stick, and colorscorrespond to the fold change in affinity or cooperativity shown in (A and C).

CaSR Activation Networks 623

at ASPE

T Journals on July 8, 2020

molpharm

.aspetjournals.orgD

ownloaded from

There were, however, subtle differences in the effects ofsome mutations on NPS R568 pKB when compared withcinacalcet. For instance, W8186.50A decreased, whereasE767ECL2A, A7725.39V, or V833ECL3A increased, cinacalcetpKB (Leach et al., 2016); in contrast, these mutations had nosignificant effect on NPS R568 pKB (Fig. 3A; Table 1). Tobetter understand the differences in the pKB effects of thesemutations, we constructed a homologymodel of the CaSR 7TMdomain based on themGlu5 7TM structure (Christopher et al.,2015) and interrogated this model bound to NPS R568in silico. Since the NPS R568 core consists of an arylalkyl-amine, whose secondary amine is predicted to interact directlywith E8377.32, the distance required for hydrogen bondingbetween this residue and the secondary amine of NPS R568was maintained during the docking process. Unsurprisingly,this produced docking poses for NPS R568 that were consis-tent with previously published predictions for cinacalcet andother arylalkylamine-based modulators (Leach et al., 2016)(Fig. 3B).Close inspection of docking poses for NPS R568 in the CaSR

homology model explained the observed differences in muta-tional effects on cinacalcet versus NPS R568 pKB. First, ourpredicted docking poses oriented themethoxyphenyl moiety ofNPS R568 toward the hydrophobic-aromatic network thatsurrounds the napthyl group of cinacalcet (F6843.36, F7755.42,

L7765.43, W8186.50, and I8417.36), but crucially, the smallerandmore hydrophilic NPS R568 moiety does not interact withW8186.50 (Fig. 3B; Supplemental Fig. 3). This change inbinding mechanism likely underpins why W8186.50A doesnot influence NPS R568 pKB. Second, the predicted NPSR568 pose places the chloro-phenyl group of NPS R568 in aposition identical to the tri-fluoromethyl-phenyl group of

cinacalcet, but the NPS R568 moiety occupies appreciablyless chemical space. Indeed, the increase in cinacalcet pKB

observed at the E767ECL2A and V833ECL3A mutants is likelydue to an increase in the volume of the top of the bindingpocket, which is more favorable for cinacalcet but makes nodifference to the binding of the smaller NPS R568 (Supple-mental Fig. 3). Collectively, these data suggest that NPS R568and cinacalcet bind to an overlapping binding site; however,due to key differences in their chemical topology, there aresubtle distinctions in the contribution of select residues to thebinding of the two PAMs.Identification of Residues that Mediate Cooperativ-

ity between Ca21o and NPS R568. By applying an opera-

tional model of allosterism to our data, we also revealed nineresidues within the CaSR 7TM bundle that are involved inNPS R568 cooperativity (logab) but not affinity (pKB), withmajor contributions from ECL1 and 2 as well as TMs 6 and7 (WT and representative mutants are shown in Fig. 2,quantification is shown in Fig. 3C). In contrast, five mutationsinfluenced both affinity and cooperativity; F6882.56A de-creased pKB and logab, whereas F8216.53A, E8377.32D,Y8256.42A, or A8447.39V had differential effects on the twoparameters (Fig. 3, A and C; Table 1). Thus, residues withinthe binding pocket differentially contribute to NPS R568affinity versus cooperativity with Ca21o .Our previous structure-function analysis at the CaSR

identified residues that contributed to the transmission ofpositive cooperativity between Ca21o and cinacalcet (Leachet al., 2016). Comparisons between cooperativity determi-nants for cinacalcet and NPS R568 revealed key similaritiesand differences. Five mutations (E767ECL2A, V8176.49A,Y8256.57A, A8447.39V, and L8487.43A) significantly reduced

TABLE 1Functional affinity (pKB) and cooperativity (logab) parameters for NPS R568 determined in Ca2+imobilization assays using an interaction paradigmData are mean 6 S.E.M. and 95% confidence interval from the indicated number of independent experiments (n).

NPS R568

pKB (95% CI) Logab (95% CI) (ab) n

WT 5.99 6 0.05 (5.89–6.09) 0.55 6 0.01 (0.52–0.58) (3.5) 17F6121.39A 6.39 6 0.15 (6.03–6.75) 0.35 6 0.04a (0.28–0.43) (2.2) 4A6151.42V 6.22 6 0.13 (5.93–6.49) 0.45 6 0.04 (0.38–0.53) (2.8) 6F6682.56A ,5 ND 4G670ECL1A 6.39 6 0.08 (6.23–6.55) 0.09 6 0.04a (20.07–0.25) (1.2) 4F6843.36A 5.28 6 0.18a (4.79–5.63) 0.35 6 0.05a (0.28–0.51) (2.2) 4F6883.40A 5.40 6 0.16a (5.00–5.73) 0.52 6 0.05 (0.44–0.65) (3.3) 5E767ECL2A 6.36 6 0.22 (5.93–6.77) 0.25 6 0.03a (0.18–0.33) (1.8) 3A7725.39V 6.33 6 0.08 (6.17–6.48) 0.38 6 0.02a (0.35–0.42) (2.4) 4F7755.42A 6.11 6 0.07 (5.95–6.27) 0.39 6 0.02a (0.36–0.42) (2.5) 4L7765.43A 6.40 6 0.09 (6.20–6.58) 0.46 6 0.03 (0.40–0.51) (2.9) 4V8176.49A 5.89 6 0.26 (5.29–6.38) 0.37 6 0.05a (0.27–0.50) (2.3) 4W8186.50A 5.88 6 0.05 (5.77–5.98) 0.55 6 0.01 (0.52–0.58) (3.5) 5F8216.53A 7.14 6 0.18a (6.80–7.58) 0.18 6 0.02a (0.14–0.22) (1.5) 4Y8256.57A 6.88 6 0.11a,b (6.65–7.11) 0.07 6 0.02a,c (0.04–0.10) (1.2) 4K831ECL3A 5.60 6 0.15 (5.26–5.91) 0.60 6 0.05 (0.49–0.71) (4.0) 4V833ECL3A 6.07 6 0.11 (5.77–6.33) 0.35 6 0.02a (0.31–0.40) (2.2) 4E8377.32D 7.13 6 0.16a (6.85–7.52) 0.26 6 0.03a (0.22–0.32) (1.9) 4E8377.32I ,5 ND 7A8407.35V 6.31 6 0.13 (6.05–6.56) 0.33 6 0.03a (0.28–0.39) (2.1) 4I8417.36A NA NA 4A8447.39V 6.69 6 0.09a (6.49–6.90) 0.05 6 0.09a (20.17–0.26) (1.1) 5L8487.43A 6.04 6 0.13 (5.79–6.28) 0.29 6 0.02a (0.25–0.32) (1.9) 4

CI, confidence interval; NA, negligible PAM activity; ND, not determined due to weak cooperativity or low affinity.asignificantly different to WT (P , 0.05, one-way ANOVA with Dunnett’s multiple comparisons post-test).bpKB determined in interaction studies between AC265347 and NPS R568 (n = 3).cpKB was fixed to 6.88 determined in interaction studies between AC265347 and NPS R568.

624 Keller et al.

at ASPE

T Journals on July 8, 2020

molpharm

.aspetjournals.orgD

ownloaded from

the cooperativity of both NPS R568 and cinacalcet, suggestingcontributions to a common network that transmits positivecooperativity by structurally similar PAMs (Fig. 3, C and D).In contrast, six of the mutations that specifically reduce NPSR568 cooperativity with Ca21o without impacting affinity(F6121.39A, G670ECL1A, A7725.39V, F7755.42A, V833ECL3A,and A8407.35V) did not influence cinacalcet cooperativity.Furthermore, A6151.42V, W8186.50A, and K831ECL3A, whichwe previously reported to influence cinacalcet cooperativity,had no significant effect on the cooperativity between NPSR568 and Ca21o , demonstrating that the contribution of theseresidues to the transmission of positive cooperativity is ligandspecific. Collectively, our data reveal differences in thenetwork of residues that transmit cooperativity mediated bystructurally similar PAMs, indicating that distinct conforma-tional states are favored by cinacalcet and NPS R568.Identification of Global and Ligand-Specific Activa-

tion Mechanisms at the CaSR. Having identified the NPSR568 binding site and differentiated ligand-receptor interac-tions governing affinity and cooperativity, we next sought tocompare activation mechanisms used by different CaSRactivators. To do so, we first validated the ability of theoperational model of allosterism with ambient agonist (eq. 1)to distinguish between PAMs (e.g., NPS R568) and PAM-agonists (e.g., AC265347). When applied to Ca21o versus NPSR568 interaction data, we found that the operational model ofallosterism with ambient agonist fit the data best when thePAM was presumed to have no efficacy (i.e., the best fit valuefor tB did not differ from zero, F test; P , 0.001) (Fig. 2A;Tables 1 and 2). This agreed with our previous study showingthat cinacalcet, NPS R568, and other structurally related

PAMs demonstrate negligible intrinsic agonism (Cook et al.,2015). In contrast, and also consistent with our previous study(Cook et al., 2015), the interaction between Ca21o and thePAM-agonist AC265347 were fitted best when the modelassumed AC265347 was an agonist (i.e., the best fit value fortB differed from zero, F test; P , 0.001) (Fig. 4A; Table 2).Importantly, reanalysis of these previously reported data witheq. 1 yielded similar estimates of AC265347 pKB and logab(Supplemental Table 1). Taken together, these findingsdemonstrate that our operational model of allosterism withambient agonist can differentiate between two PAMs withdifferent agonist activity. Therefore, we next assessed theimpact of mutations on allosteric modulator efficacy, tB,independently of positive cooperativity.Alanine substitution of F6843.36 or F6883.40 was previously

found to reduce Ca21o potency (Leach et al., 2016), and in thecurrent study also abolished AC265347-mediated Ca21i mobi-lization in the presence of vehicle (while positive cooperativitywith Ca21o was retained) (Fig. 4B; Supplemental Fig. 1; Tables1 and 2). Using a tetracycline-inducible cell line, we titratedF6843.36 and F6883.40 receptor expression levels to establishthe Ca21o pEC50 under conditions where it should more closelyreflect its affinity (i.e., in the absence of receptor reserve)(Supplemental Fig. 4). With reduced expression levels, Ca21opEC50 values at the F684

3.36A (2.906 0.07) or F6883.40A (2.956 0.03)mutants were not significantly different fromWT (3.066 0.04), indicating that the reduction in Ca21o potency at themutants versusWT under high expression conditions was dueto reduced Ca21o efficacy, not affinity. In contrast, F6121.39A,L7765.43A, or V8176.49A increased Ca21o potency (Leach et al.,2016) and AC265347 efficacy (Table 2). The changes in Ca21o

TABLE 2Efficacy (tB) of NPS R568 and AC265347 determined in Ca2+i mobilization assays using an interactionparadigmData are mean 6 S.E.M. and 95% confidence intervals from the indicated number of independent experiments (n).

NPS R568 AC265347

LogtB (95% CI) (tB) n LogtB (95% CI) (tB) n

WT ND 17 20.22 6 0.09 (20.45 to 20.08) (0.6) 8F6121.39A ND 4 0.12 6 0.09a (20.10 to 0.44) (1.3) 4A6151.42V ND 6 0.04 6 0.03 (20.01 to 0.09) (1.1) 4F6682.56A ND 4 ND 4G670ECL1A 20.25 6 0.06 (20.41 to 20.15) (0.6) 4 20.15 6 0.09 (20.39 to 0.02) (0.7) 4F6843.36A ND 4 ND 5F6883.40A ND 5 ND 4E767ECL2A ND 3 0.05 6 0.03 (20.02 to 0.12) (1.1) 4A7725.39V ND 4 ND 4F7755.42A ND 4 20.26 6 0.07 (20.48 to 0.14) (0.5) 7L7765.43A ND 4 0.19 6 0.12a (20.11 to 0.32) (1.5) 6V8176.49A ND 4 0.33 6 0.04a,b (0.26–0.45) (2.1) 5W8186.50A ND 5 ND 4F8216.53A ND 4 ND 4Y8256.57A ND 4 ND 4K831ECL3A ND 4 20.05 6 0.03 (20.10 to 0.00) (0.9) 5V833ECL3A ND 4 ND 5E8377.32D ND 4 0.09 6 0.12 (20.18 to 0.28) (1.2) 4E8377.32I NA 7 NA 5A8407.35V ND 4 ND 7I8417.36A NA 4 ND 4A8447.39V 20.22 6 0.07 (20.45 to 20.11) (0.6) 5 20.12 6 0.03 (20.14 to 20.09) (0.8) 6L8487.43A ND 4 20.17 6 0.07 (20.35 to 20.07) (0.7) 4

CI, confidence interval; NA, negligible PAM activity; ND, not determined due to negligible efficacy.aSignificantly different from WT (P , 0.05, one-way ANOVA with Dunnett’s multiple comparisons post-test).bTo account for a reduction in cell surface expression caused by V8176.49A compared with the WT CaSR (Leach et al.,

2016), logtB was normalized to determine what it would be if this mutant was expressed at the same level as the WTreceptor (Gregory et al., 2010).

CaSR Activation Networks 625

at ASPE

T Journals on July 8, 2020

molpharm

.aspetjournals.orgD

ownloaded from

potency and AC265347 efficacy at the F6843.36A, F6883.40A,F6121.39, L7765.43, or V8176.49A mutants under conditionswhere WT and mutant cell surface expression levels arecomparable (Leach et al., 2016) suggests these residues areimportant for the transmission of agonism imparted by bothCa21o and AC265347.In contrast, W8186.50A, F6682.56A, A7725.39V, F8216.53A,

Y8256.57A, V833ECL3A, A8407.35V, or I8417.36A did not reduceCa21o potency (several increased it), or with the exception ofF6682.56A did not reduce cell surface expression (Leach et al.,2016), but they all abolished AC265347 efficacy (Fig. 4D;Table 2). Qualitatively, AC265347 enhanced Ca21i mobilizationin the presence of vehicle alone at certain mutations (see, forexample, Fig. 4D); however, our analysis is consistentwith thesebaseline increases being attributable to positive modulation ofambient cations in the buffer rather than intrinsic AC265347efficacy (Supplemental Table 1). Therefore, these eight residuesare critical for mediating agonism of AC265347, but not Ca21o .Intriguingly, analysis of NPS R568 interactions with Ca21o

revealed two mutations (G670AECL1A and A8447.39V) wherethere was a gain in allosteric agonism, as demonstrated by asignificant increase in tB, but neither mutation enhancedAC265347 efficacy (Table 2). Combined with a comparison ofmutations that influence Ca21o potency (Supplemental Fig. 5)versus AC265347 efficacy (Fig. 5; Table 2), our findingsindicate that some 7TM and ECL residues play a global rolein agonism imparted by all classes of agonist (i.e., orthostericand allosteric), whereas others contribute to ligand-specificreceptor activation.The finding that the G670A and A844V mutations

significantly reduced NPS R568 cooperativity yet convertedNPS R568 to a PAM-agonist was intriguing. Therefore, wenext sought to compare the global and AC265347-specific

activation residues with the AC265347 cooperativity net-works previously established (Leach et al., 2016) to determinewhether there were also distinctions between cooperativityversus agonism networks used by other PAMs. Twomutations(F6121.39A and L7765.43A) that increased AC265347 efficacy(Fig. 5, A and B) did not change positive cooperativity,whereas V8176.49A increased AC265347 efficacy while de-creasing cooperativity. In contrast, the loss in AC265347efficacy at F6682.56A was opposed by increases in coopera-tivity. Four mutations that decreased AC265347 cooperativity(A6151.42V, E767ECL2A, K831ECL3A, and A8447.39V) had noeffect on AC265347 efficacy. Collectively, these data revealdifferential contributions of residues within the 7TM bundleand ECLs to efficacy versus cooperativity.CaSR PAMs Enhance the Affinity of Orthosteric

Agonists. To confirm mutational effects on PAM efficacypredicted from our analysis of interaction data, we assessedthe ability of NPS R568 or AC265347 to stimulate Ca21imobilization in the absence of ambient Ca21o or Mg21o at WTand select mutant receptors. We evaluated G670ECL1A andA8447.39V mutations, where NPS R568 was predicted to gainagonist activity, and A8407.35V as a representative mutationthat decreased AC265347 efficacy (Table 2). Furthermore, weevaluated E8377.32D as a control, because this mutation hadno effect on the efficacy of NPS R568 or AC265347. Wherepossible, we fit agonist concentration response data to anoperational model of agonism (eq. 2) to determine agonistaffinity and efficacy at WT and/or mutant CaSRs (Table 3).In agreement with our previous evaluation of PAM agonism

(Cook et al., 2015), up to 30 mM NPS R568 did not stimulateCa21i mobilization in the absence of ambient cations (Fig. 6A);due to limited solubility, it was not possible to test higherconcentrations. Nonetheless, these findings confirm that NPS

Fig. 4. 7TM domain mutations alter orthosteric andallosteric agonist activity at the CaSR. Concentration-response curves to Ca2+o determined inCa2+i mobilizationassays in the absence and presence of AC265347 (A–D)have been published previously (Leach et al., 2016), andwere reanalyzed with an operational model of alloster-ism that considered ambient divalent cations to de-termine the effect of mutations on AC265347 efficacy(tB). WT and representative mutants are shown thatcaused a simultaneous reduction in Ca2+o potency andloss of AC265347 efficacy (F6883.40A), a simultaneousincrease in Ca2+o potency and gain in AC265347 efficacy(L7765.43A), and a selective loss of AC265347 efficacy(A8407.35V). Data aremean + S.E.M. pooled from at leastfour separate experiments performed in duplicate.Curves through the data points are the best fit of thedata to eq. 1. Vehicle (buffer) is plotted as the smallest[Ca2+o ] in accordance with the logarithmic scale.

626 Keller et al.

at ASPE

T Journals on July 8, 2020

molpharm

.aspetjournals.orgD

ownloaded from

R568 exhibits negligible agonism at the CaSR. In contrast toWT CaSR, NPS R568 stimulated concentration-dependentCa21i mobilization in the absence of ambient cations atG670ECL1A and A8447.39V (Fig. 6, B and C), suggestive ofincreased NPS R568 efficacy, as predicted by our operationalmodel analysis that incorporated the presence of ambientcations (Table 2). However, the estimated NPS R568 pKB atthe G670ECL2A mutant was significantly lower in the absenceof ambient cations (Table 3) than in their presence (Table 1)(P , 0.05, two-way ANOVA with Sidak’s multiple compari-sons). A similar trend toward a lower NPS R568 pKB in theabsence of ambient cations was also observed at the A8447.39Vmutant, but this did not reach significance. At E8377.32D andA8407.35V, NPS R568 exhibited negligible agonist activity,similar to WT CaSR and as predicted from our operationalmodel of allosterism analysis.In contrast to NPS R568, AC265347 stimulated relatively

robust Ca21i mobilization at the WT CaSR in the absence of

ambient cations (Fig. 6A; Table 3), indicative of intrinsicagonism. AC265347 was also an agonist at the G670ECL2A,E8377.32D, and A8447.39V mutants in the absence of ambientcations, although there was a small but significant decrease inits efficacy at A8447.39V (Fig. 6; Table 3). Consistent with ouroperational model of allosterism analysis, AC265347 efficacywas reduced at the A8407.35V mutant.Similar to NPS R568 at WT CaSR and G670ECL2A,

A8447.39V, E8377.32D, or A8407.35V mutants, AC265347bound with lower affinity in the absence of ambient cationsthan in their presence (P , 0.05, two-way ANOVA withSidak’s multiple comparisons). The difference in affinitiessuggests that CaSR PAMs bind preferentially when anorthosteric agonist simultaneously occupies the receptor.Since cooperative allosteric effects on affinity are reciprocal(and, therefore, affect both agonist and modulator) (Mayet al., 2007), the elevated PAM affinity in the presence ofambient cations is expected if the PAM also enhances the

Fig. 5. 7TMandECLmutations alter AC265347 efficacy at theCaSR. (A) Previously published concentration-response curvesto Ca2+o in the absence and presence of AC265347 determined inCa2+i mobilization assays (Leach et al., 2016) were reanalyzedwith an operational model of allosterism that accounted forambient divalent cations (eq. 1) to determine the change inAC265347 efficacy (DtB) at the mutant CaSRs in comparisonwith the WT receptor. White bars represent no significantchange in tB. Colored bars that sit above zero represent anincrease in tB. A significant 1.6–2.5-fold increase in tB isrepresented by the purple bar and a greater than 2.5-foldincrease is represented by blue bars. ND, not determined due tonegligible efficacy; NA, no PAMactivity. Data are theDtB + S.E.M. calculated from WT and mutant tB values and experimentnumbers shown in Table 2; * demotes significant difference incomparison with WT (P , 0.05, one-way ANOVA with Dun-nett’s multiple comparisons post-test). Efficacy residues areshown on a CaSR 7TM domain homology model (B), along withthe predicted AC265347 pose from docking studies. Blue andpurple colors correspond to the fold change in efficacy shown in(A), whereas red corresponds to a complete loss in AC265347efficacy.

CaSR Activation Networks 627

at ASPE

T Journals on July 8, 2020

molpharm

.aspetjournals.orgD

ownloaded from

affinity of the ambient cations. Thus, NPS R568 andAC265347 likely act, at least in part, by potentiating theaffinity of divalent cations. This could occur if the PAMsstabilized the closed form of the VFT via allosteric interac-tions between the 7TM and VFT domains.

DiscussionThe current study has used a structure-function approach

combined with analytical pharmacology to probe CaSR 7TMdomain and ECL residues that contribute to receptor

activation and positive allosteric modulation. Importantly,by incorporating the presence of ambient agonist into anoperational model of allosterism, wewere able to delineate thecontribution of residues to PAM intrinsic efficacy (agonism)independently of positive cooperativity (potentiation of theorthosteric agonist) and affinity.We found both ligand-specificand global activation networks within the CaSR ECLs and7TM bundle. The fact that mutations in these regions of thereceptor have substantial effects on both orthosteric andallosteric ligands highlights the pivotal role that the 7TMdomain and ECLs play in transmitting activating stimuli,

Fig. 6. CaSR PAMs bind with lower affinity in the absencevs. presence of ambient divalent cations, and exhibitdifferential efficacy at the 7TM domain and ECL mutants.Concentration-response curves to Ca2+o , NPS R568, andAC265347 were determined in Ca2+i mobilization assays inthe absence of ambient divalent cations to evaluate PAMefficacy (A–E). Representative mutants were chosen basedon their ability to provoke NPS R568 efficacy (G670ECL2Aand A8447.39V), reduce AC265347 efficacy (A8407.35V), orhave no effect on PAM efficacy (E8377.32D). Data aremean +S.E.M. pooled from at least four separate experimentsperformed in duplicate. Curves are the best fit of the datato eq. 2. Vehicle (buffer) is plotted as the smallest [Ca2+o ] inaccordance with the logarithmic scale.

TABLE 3Functional affinity (pKB) and efficacy (logtB) parameters for NPS R568 and AC265347 determined in Ca2+i mobilization assays in the absence ofambient divalent cationsData are mean 6 S.E.M. and 95% confidence intervals from the indicated number of independent experiments (n).

NPS R568 AC265347

pKB (95% CI) LogtB (95% CI) (tB) n pKB (95% CI) LogtB (95% CI) (tB) n

WT ND ND 8 5.43 6 0.10a (5.23–5.63) 0.05 6 0.10 (0.01–0.11) (1.1) 8G670ECL1A 5.71 6 0.18a (5.28–6.13) 20.19 6 0.04 (20.27 to 20.11) (0.6) 7 6.13 6 0.08b (5.96–6.31) 0.07 6 0.03 (0.04–0.11) (1.2) 7E8377.32D ND ND 8 5.90 6 0.14a,b (5.60–6.19) 20.03 6 0.03 (20.10 to 0.03) (0.9) 8A8407.35V ND ND 7 ND ND 7A8447.39V 6.48 6 0.32 (5.78–7.21) 20.27 6 0.06 (20.42 to 20.17) (0.5) 6 6.89 6 0.20b (6.51–7.31) 20.16 6 0.04b (20.25 to 20.09) (0.7) 6

CI, confidence interval; ND, not determined due to negligible efficacy and/or low affinity.aSignificantly lower than in the presence of ambient divalent cations (P , 0.05, two-way ANOVA with Sidak’s comparisons post-test).bSignificantly different from WT (P , 0.05, one-way ANOVA with Dunnett’s multiple comparisons post-test).

628 Keller et al.

at ASPE

T Journals on July 8, 2020

molpharm

.aspetjournals.orgD

ownloaded from

irrespective of the location of the ligand binding site. Impor-tantly, our ability to delineate mutation effects on coopera-tivity and efficacy reveals that positive cooperativity appearsto be mediated in part by residues that are distinct to thosethat transmit allosteric agonism. For instance, A6151.42,E767ECL2, and K831ECL3 all contributed to the transmissionof AC265347 cooperativity yet played no role in AC265347efficacy. Conversely, F6121.39 and L7765.43 constrainedAC265347 efficacy but not positive cooperativity. In otherinstances, there were opposing contributions of residues tocooperativity versus efficacy, as was observed for G670ECL2

and A8447.39; mutation of these residues decreased NPS R568cooperativity yet convertedNPSR568 to an agonist. Similarly,V8176.49A increased AC265347 efficacy while decreasingcooperativity, where the opposite was true for F6682.56A.Using mutagenesis to dissect the contribution of 7TM and

ECL residues toNPSR568 affinity, we confirm thatNPSR568binds to an extended cavity in the CaSR’s 7TM domain, muchlike cinacalcet, with key interactions between F6682.56,F6843.36, F6883.40, E8377.32, and I8417.36. We did, however,identify some subtle differences in the binding of NPS R568and cinacalcet, likely attributed to their different terminalaromatic groups, permitting cinacalcet to occupy more spaceat the top and bottom of the binding pocket. Despite this, andthe identification of a number of residues that differentiallycontribute to NPS R568 and cinacalcet cooperativity, thesetwo PAMs from the arylalkylamine family are pharmacolog-ically very similar (Cook et al., 2015). The differences in NPSR568 and cinacalcet SAR appear to be overcome by sharedcontributions of E767ECL2, V8176.49, Y8256.57, A8447.39, andL8487.43 to their transmission of positive cooperativity. Im-portantly, E767ECL2, V8176.49, and A8447.39 are also criticalfor positive cooperativity mediated by the structurally distinctPAM-agonist AC265347, suggesting that these residues maycontribute to a global allosteric activation network. Further-more, mutation of these residues also increases Ca21o potency,suggesting a reduction in the energy barrier to transition to anactive CaSR 7TM domain that mirrors the effects of PAMs. Infact, greater than 70% of the 7TM/ECL mutations studiedherein increase Ca21o potency (Leach et al., 2016), supportingthe idea that substitution of residues in the allosteric bindingcavity lowers the activation barrier for Ca21o . Mutationsthat reduce Ca21o potency (e.g., F6843.36A, F6883.40A, andE8377.32I) were at residues critical to the binding of allostericdrugs. Collectively, the influence of these 7TM/ECL residueson both orthosteric agonist activity and small-molecule bind-ing and agonism indicates that the 7TM allosteric binding sitehas evolutionary importance and raises the possibility that anendogenous allosteric modulator may act via this site.With regards to mutational effects on agonism, it was of

particular note that aside from L7765.43 and V8176.49, whichwere important for the efficacy of both Ca21o and AC265347,the majority of residues differentially contributed to orthos-teric versus allosteric efficacy. This was demonstrated by theopposing effect of many mutations, which increased Ca21opotency yet decreased AC265347 efficacy. We cannot excludethe possibility that changes in Ca21o potency were due toaltered Ca21o affinity, but we did show this was not the case forF6843.36A and F6883.40A mutations (Supplemental Fig. 4).With the exception of E8377.32, which we postulate may bindCa21o , it is unlikely that mutation of other residues in thispocket significantly alter Ca21o affinity, given that four key

Ca21o binding sites are located in the receptor’s VFT domain(Geng et al., 2016). Thus, our data support the existence ofglobal and ligand-specific activation mechanisms in theCaSR’s 7TM and ECL regions.Another key finding of this study is the demonstration that

CaSR PAMs act by potentiating the affinity of Ca21o . This is incontrast to other class C GPCR PAMs, which enhanceorthosteric agonist-mediated efficacy but do not appear topotentiate their binding (Gregory et al., 2012). The significantloss in PAM affinity in the absence of ambient cations versusin their presence is reminiscent of the reciprocal interactionbetween Ca21o and L-amino acids (Conigrave et al., 2000,2004); structural studies suggest both ligands are bound to theCaSR’s VFT domain to achieve a fully active receptor confor-mation (Geng et al., 2016; Zhang et al., 2016). However, unlikeL-amino acids, our current and previous studies (Leach et al.,2016) demonstrate that small-molecule PAMs bind in anextended CaSR 7TM domain cavity.Our findings have important implications for future drug

discovery efforts at the CaSR. Cinacalcet use is restricted dueto an adverse risk of hypocalcemia (Chonchol et al., 2009), inpart caused by oversuppression of parathyroid hormonesecretion along with stimulation of thyroid CaSRs andconsequent calcitonin secretion, and/or suppression of Ca21oreabsorption via activation of renal CaSRs (reviewed in Leachet al., 2014, 2015). There is consequently continued interest indeveloping CaSR activators with greater tissue selectivity.Indeed, weak (partial) agonists or PAMs would be expected tostimulate robust pharmacological responses only in tissueswhere strong stimulus-response coupling exists, such as highCaSR-expressing parathyroid cells. To date, no small-molecule CaSR agonists devoid of PAM activity have beenidentified, and there have been no reported efforts to examinethe physiologic consequences of PAMs with different degreesof cooperativity. Furthermore, as demonstrated by AC265347,somePAMs also exhibit agonism,whichmay heighten adverseeffects by further amplifying receptor signaling. A greaterappreciation of the drivers of CaSR agonism versus allostericpotentiationmay afford an opportunity to rationally tune in ortune out PAM and agonist activity.In conclusion, the current study provides new insight into

7TM and ECL CaSR activation mechanisms. In the future,this information could facilitate rational structure-based de-sign of novel CaSR activators whose pharmacological andconsequent therapeutic properties could be optimized by fine-tuning their PAM and/or agonist activity.

Authorship Contributions

Participated in research design: Gregory, Leach.Conducted experiments: Keller, Kufareva, Josephs, Diao, Mai,

Gregory, Leach.Performed data analysis: Gregory, Leach.Wrote or contributed to the writing of the manuscript: Keller,

Kufareva, Conigrave, Christopoulos, Gregory, Leach.

References

Abagyan R and Totrov M (1994) Biased probability Monte Carlo conformationalsearches and electrostatic calculations for peptides and proteins. J Mol Biol 235:983–1002.

Aurelio L, Valant C, Flynn BL, Sexton PM, Christopoulos A, and Scammells PJ(2009) Allosteric modulators of the adenosine A1 receptor: synthesis and phar-macological evaluation of 4-substituted 2-amino-3-benzoylthiophenes. J Med Chem52:4543–4547.

Bu L, Michino M, Wolf RM, and Brooks CL, III (2008) Improved model building andassessment of the calcium-sensing receptor transmembrane domain. Proteins 71:215–226.

CaSR Activation Networks 629

at ASPE

T Journals on July 8, 2020

molpharm

.aspetjournals.orgD

ownloaded from

Chonchol M, Locatelli F, Abboud HE, Charytan C, de Francisco AL, Jolly S, Kaplan M,Roger SD, Sarkar S, Albizem MB, et al. (2009) A randomized, double-blind, placebo-controlled study to assess the efficacy and safety of cinacalcet HCl in participantswith CKD not receiving dialysis. Am J Kidney Dis 53:197–207.

Christopher JA, Aves SJ, Bennett KA, Doré AS, Errey JC, Jazayeri A, Marshall FH,Okrasa K, Serrano-Vega MJ, Tehan BG, et al. (2015) Fragment and structure-baseddrug discovery for a class C GPCR: discovery of the mGlu5 negative allo-steric modulator HTL14242 (3-chloro-5-[6-(5-fluoropyridin-2-yl)pyrimidin-4-yl]-benzonitrile). J Med Chem 58:6653–6664.

Christopoulos A (1998) Assessing the distribution of parameters in models of ligand-receptor interaction: to log or not to log. Trends Pharmacol Sci 19:351–357.

Conigrave AD, Mun HC, Delbridge L, Quinn SJ, Wilkinson M, and Brown EM (2004)L-amino acids regulate parathyroid hormone secretion. J Biol Chem 279:38151–38159.

Conigrave AD, Quinn SJ, and Brown EM (2000) L-amino acid sensing by the ex-tracellular Ca21-sensing receptor. Proc Natl Acad Sci USA 97:4814–4819.

Cook AE, Mistry SN, Gregory KJ, Furness SG, Sexton PM, Scammells PJ, ConigraveAD, Christopoulos A, and Leach K (2015) Biased allosteric modulation at the CaSreceptor engendered by structurally diverse calcimimetics. Br J Pharmacol 172:185–200.

Davey AE, Leach K, Valant C, Conigrave AD, Sexton PM, and Christopoulos A (2012)Positive and negative allosteric modulators promote biased signaling at thecalcium-sensing receptor. Endocrinology 153:1232–1241.

de Graaf C, Song G, Cao C, Zhao Q, Wang MW, Wu B, and Stevens RC (2017)Extending the structural view of class B GPCRs. Trends Biochem Sci 42:946–960.

Doré AS, Okrasa K, Patel JC, Serrano-Vega M, Bennett K, Cooke RM, Errey JC,Jazayeri A, Khan S, Tehan B, et al. (2014) Structure of class C GPCR metabotropicglutamate receptor 5 transmembrane domain. Nature 511:557–562.

Geng Y, Mosyak L, Kurinov I, Zuo H, Sturchler E, Cheng TC, Subramanyam P,Brown AP, Brennan SC, Mun HC, et al. (2016) Structural mechanism of ligandactivation in human calcium-sensing receptor. eLife 5:e13662.

Gregory KJ, Hall NE, Tobin AB, Sexton PM, and Christopoulos A (2010) Identifi-cation of orthosteric and allosteric site mutations in M2 muscarinic acetylcholinereceptors that contribute to ligand-selective signaling bias. J Biol Chem 285:7459–7474.

Gregory KJ, Nguyen ED, Malosh C, Mendenhall JL, Zic JZ, Bates BS, Noetzel MJ,Squire EF, Turner EM, Rook JM, et al. (2014) Identification of specific ligand-receptor interactions that govern binding and cooperativity of diverse modulatorsto a common metabotropic glutamate receptor 5 allosteric site. ACS Chem Neurosci5:282–295.

Gregory KJ, Nguyen ED, Reiff SD, Squire EF, Stauffer SR, Lindsley CW, Meiler J,and Conn PJ (2013) Probing the metabotropic glutamate receptor 5 (mGlu5) positiveallosteric modulator (PAM) binding pocket: discovery of point mutations that en-gender a “molecular switch” in PAM pharmacology. Mol Pharmacol 83:991–1006.

Gregory KJ, Noetzel MJ, Rook JM, Vinson PN, Stauffer SR, Rodriguez AL, EmmitteKA, Zhou Y, Chun AC, Felts AS, et al. (2012) Investigating metabotropic glutamatereceptor 5 allosteric modulator cooperativity, affinity, and agonism: enrichingstructure-function studies and structure-activity relationships. Mol Pharmacol 82:860–875.

Langmead CJ, Fry VA, Forbes IT, Branch CL, Christopoulos A, Wood MD,and Herdon HJ (2006) Probing the molecular mechanism of interaction between4-n-butyl-1-[4-(2-methylphenyl)-4-oxo-1-butyl]-piperidine (AC-42) and the musca-rinic M1 receptor: direct pharmacological evidence that AC-42 is an allosteric ag-onist. Mol Pharmacol 69:236–246.

Leach K, Conigrave AD, Sexton PM, and Christopoulos A (2015) Towards tissue-specific pharmacology: insights from the calcium-sensing receptor as a paradigmfor GPCR (patho)physiological bias. Trends Pharmacol Sci 36:215–225.

Leach K and Gregory KJ (2017) Molecular insights into allosteric modulation of classC G protein-coupled receptors. Pharmacol Res 116:105–118.

Leach K, Gregory KJ, Kufareva I, Khajehali E, Cook AE, Abagyan R, Conigrave AD,Sexton PM, and Christopoulos A (2016) Towards a structural understanding ofallosteric drugs at the human calcium-sensing receptor. Cell Res 26:574–592.

Leach K, Loiacono RE, Felder CC, McKinzie DL, Mogg A, Shaw DB, Sexton PM,and Christopoulos A (2010) Molecular mechanisms of action and in vivo validationof an M4 muscarinic acetylcholine receptor allosteric modulator with potentialantipsychotic properties. Neuropsychopharmacology 35:855–869.

Leach K, Sexton PM, and Christopoulos A (2007) Allosteric GPCR modulators: takingadvantage of permissive receptor pharmacology. Trends Pharmacol Sci 28:382–389.

Leach K, Sexton PM, Christopoulos A, and Conigrave AD (2014) Engendering biasedsignalling from the calcium-sensing receptor for the pharmacotherapy of diversedisorders. Br J Pharmacol 171:1142–1155.

Leach K, Wen A, Cook AE, Sexton PM, Conigrave AD, and Christopoulos A (2013)Impact of clinically relevant mutations on the pharmacoregulation and signalingbias of the calcium-sensing receptor by positive and negative allosteric modulators.Endocrinology 154:1105–1116.

Manglik A and Kruse AC (2017) Structural basis for G protein-coupled receptoractivation. Biochemistry 56:5628–5634.

May LT, Leach K, Sexton PM, and Christopoulos A (2007) Allosteric modulation of Gprotein-coupled receptors. Annu Rev Pharmacol Toxicol 47:1–51.

Mayr B, Schnabel D, Dörr HG, and Schöfl C (2016) Genetics in endocrinology: gainand loss of function mutations of the calcium-sensing receptor and associatedproteins: current treatment concepts. Eur J Endocrinol 174:R189–R208.

Miedlich SU, Gama L, Seuwen K, Wolf RM, and Breitwieser GE (2004) Homologymodeling of the transmembrane domain of the human calcium sensing receptorand localization of an allosteric binding site. J Biol Chem 279:7254–7263.

Motulsky H and Christopoulos A (2004) Fitting Models to Biological Data UsingLinear and Nonlinear Regression. A Practical Guide to Curve Fitting, OxfordUniversity Press, New York.

Nemeth EF and Goodman WG (2016) Calcimimetic and calcilytic drugs: feats, flops,and futures. Calcif Tissue Int 98:341–358.

Nemeth EF, Steffey ME, Hammerland LG, Hung BC, Van Wagenen BC, DelMar EG,and Balandrin MF (1998) Calcimimetics with potent and selective activity on theparathyroid calcium receptor. Proc Natl Acad Sci USA 95:4040–4045.

Petrel C, Kessler A, Dauban P, Dodd RH, Rognan D, and Ruat M (2004) Positive andnegative allosteric modulators of the Ca21-sensing receptor interact within over-lapping but not identical binding sites in the transmembrane domain. J Biol Chem279:18990–18997.

Petrel C, Kessler A, Maslah F, Dauban P, Dodd RH, Rognan D, and Ruat M (2003)Modeling and mutagenesis of the binding site of Calhex 231, a novel negativeallosteric modulator of the extracellular Ca21-sensing receptor. J Biol Chem 278:49487–49494.

Ray K and Northup J (2002) Evidence for distinct cation and calcimimetic compound(NPS 568) recognition domains in the transmembrane regions of the human Ca21

receptor. J Biol Chem 277:18908–18913.Zhang C, Zhang T, Zou J, Miller CL, Gorkhali R, Yang JY, Schilmiller A, Wang S,Huang K, Brown EM, et al. (2016) Structural basis for regulation of humancalcium-sensing receptor by magnesium ions and an unexpected tryptophan de-rivative co-agonist. Sci Adv 2:e1600241.

Address correspondence to: Katie Leach, Drug Discovery Biology, MonashInstitute of Pharmaceutical Sciences and Department of Pharmacology,381 Royal Parade, Parkville, VIC, Australia 3052. E-mail: katie.leach@monash.edu or Karen Gregory, Drug Discovery Biology, Monash Institute ofPharmaceutical Sciences and Department of Pharmacology, 381 Royal Parade,Parkville, VIC, Australia 3052. E-mail: karen.gregory@monash.edu

630 Keller et al.

at ASPE

T Journals on July 8, 2020

molpharm

.aspetjournals.orgD

ownloaded from